Studies On Providing Scalable Qos Solutions In IP Networks

The enlargement of the Internet user community has generated the need for IP-basedapplications requiring guaranteed Quality of Service (QoS) characteristics. Efforts to pro-viding QoS for the Internet have led to two distinct approaches: the Integrated Services(IntServ) and the Differentiated Services (DiffServ). IntServ suffers from scalability prob-lem, whereas DiffServ only supports weak QoS as compared to IntServ. A careful analysisof these basic QoS schemes reveals that providing QoS in IP network need every networkelements working harmoniously under some efficient, scalable, and practicable mechanisms.The most important issues among them are the scheduling policy, admission control, QoSrouting, and the means of resource reservation.First, this paper proposes a novel Fair Aggregator for Flows Isolation(FAFI) that com-bines both Earliest Deadline First(EDF) scheduling and Worst-Case Fair Weighted FairQueuing(WF2Q). By assembling individual ?ows in EDF and forwarding aggregated traf-fic in WF2Q, FAFI overcomes disadvantage in ?ow aggregating that the end-to-end delayof an individual ?ow is related to the burstiness of other ?ows sharing its aggregated traffic.Subsequently, an end-to-end delay bound is derived in the network equipped with proposedaggregator. In addition, a performance comparison is done with previous works on the as-pects of end-to-end delay bound, network utilization, and scalability.Second, a novel Multi-Objective Genetic Algorithms for Path Selection(MOGA-PS) isproposed for the setting up of traffic trucks in the network domain. The algorithm can opti-mize multiple QoS objectives by a genetic algorithm in conjunction with concept of Paretodominance, which is capable of discovering a set of QoS-based near optimal paths within ina few iterations. The proposed algorithm is able to eliminate the traditional algorithm draw-backs of combining different and con?icting characteristics of these QoS parameters into asingle scalar objective-function, where the real intuition and logic behind the combinations is often fuzzy. In addition, simulation results also show the efficiency and scalability of thealgorithm with increasing number of network nodes.Third, a new scalable Distributed and Parallel Admission Control(DPAC) scheme forthe QoS sensitivity traffic in core-stateless networks is proposed. In this scheme, the ingressrouters perform admissibility test in a fully distributed and parallel fashion for requests bydint of our bandwidth dividing based virtual link mechanism. Then, the paper introduce anovel two phase token passing mechanism to adaptively optimize the proportion of band-width dividing among contending edge routers according to the dynamic of their traffic. Thesimulation results are very promising indicating that even under very high request load it ispossible to perform admission control and resource allocation in parallel without sufferingin terms of response time, packet loss rate, or utilization.Fourth, a new practicable QoS framework with Distributed Adaptive Resource Man-agement (DARM) is proposed. DARM provides per-?ow admission control and resourcereservation as IntServ while keeping the salability characteristic as DiffServ. In conjunctionwith a novel IPv6 ?ow label mechanism, DARM is capable of making admission controldecisions instantaneously in a fully distributed and independent manner at edge of networkwithout hop-by-hop signaling. Moreover, adaptive resource reconfiguration and distributedresource probing mechanism lead to higher global resource utilization in DARM domain.The simulation results clearly exhibit that DARM has a better overall performance compar-ing to the IntServ and DiffServ.